摘要：Understanding pests and their life cycle are a complex phenomenon due to their nonlinear relation with environmental factors and the interdependence of the environmental factors themselves. We focus our attention on ML (Machine Learning) models due to their ability to simulate non-linear phenomena effectively. We use Stacked Models, that consist of a neural network model (NN) and a time series model (TS) to simulate the life cycle of pests which varies in duration with the seasons. Several Machine Learning models were developed for predicting Helopeltis (tea pest, though impacts several other crops as well) infestation. The idea was not to depend heavily on the good scenario data, but rather which is readily available even for small holders and planters who are not too much reliant on technology to capture the data on pest infestation. Evaluating the various developed models shows that Neural Network models with better accuracies can be developed with real world data from the field. Thus, being suitable for tea gardens without too much reliance on technologies and extensive data capturing processes. This has been possible through special treatment of available data as well as ability of stacked models to provide good results in the solution space with a generally smaller set of data having lesser number of parameters.
